Automated Formative Feedback for Algorithm and Data Structure Self-Assessment

doi:10.3390/electronics14051034

Return

Automated Formative Feedback for Algorithm and Data Structure Self-Assessment

2025-03-05

Lourdes Araujo *

Fernando López-Ostenero

Laura Plaza

Juan Martínez-Romo

DOI：10.3390/electronics14051034

Original

Original request for help

Save

Abstract

En 中文

Self-evaluation empowers students to progress independently and adapt their pace according to their unique circumstances. A critical facet of self-assessment and personalized learning lies in furnishing learners with formative feedback. This feedback, dispensed following their responses to self-assessment questions, constitutes a pivotal component of formative assessment systems. We hypothesize that it is possible to generate explanations that are useful as formative feedback using different techniques depending on the type of self-assessment question under consideration. This study focuses on a subject taught in a computer science program at a Spanish distance learning university. Specifically, it delves into advanced data structures and algorithmic frameworks, which serve as overarching principles for addressing complex problems. The generation of these explanatory resources hinges on the specific nature of the question at hand, whether theoretical, practical, related to computational cost, or focused on selecting optimal algorithmic approaches. Our work encompasses a thorough analysis of each question type, coupled with tailored solutions for each scenario. To automate this process as much as possible, we leverage natural language processing techniques, incorporating advanced methods of semantic similarity. The results of the assessment of the feedback generated for a subset of theoretical questions validate the effectiveness of the proposed methods, allowing us to seamlessly integrate this feedback into the self-assessment system. According to a survey, students found the resulting tool highly useful.

Keywords:

online learning

formative feedback

natural language processing

semantic similarity

large language models